SMasks are greyscale images that get used as alpha channel for a
different image.
JPEGs in PDFs are stored as streams with /DCTDecode filters, and
we have a separate code path for loading those in the PDF renderer.
That code path just calls our JPEG decoder, which creates bitmaps
with format BGRx8888.
So when we process an SMask for such a bitmap, we have to change
the bitmap's format to BGRA8888 in addition to setting alpha values
on all pixels.
This is a hack: Ideally we'd have a CMYK Bitmap pixel format,
and we'd convert to rgb at blit time. Then we could also apply color
profiles (which for CMYK images are CMYK-based).
Also, the colors for our CMYK->RGB conversion are off for PDFs,
and we have distinct codepaths for this in Gfx::Color (for paths)
and JPEGs. So when we fix that, we'll have to fix it in two places.
But this doesn't require a lot of code and it's a huge visual
progression, so let's go with it for now.
Increases code reuse between encoder and decoder a tiny bit.
No behavior change.
(We should use the vectorized version in the future! But first,
we should learn to do the prediction in-place.)
The differencing predictor is a different way to encode pixels in TIFF
images. Every pixel is encoded as a difference with the previous column
of the image, except the first column, obviously.
This parameter is materialized by a new tag for which reading was also
implemented.
Previously, it was possible for a `TextDescriptionTagData` object with
an incorrect Macintosh ScriptCode description length to cause a buffer
overflow.
Add two new members to Font::AllowInexactMatch: Larger and Smaller.
This allows callers of FontDatabase::get to specify if they want to
prefer a larger or smaller font in the event of a tie.
Read the basic lists as spans, and use those when looking for kerning.
Kerning lookup still does bit-casting for now. As for CBLC, the data is
a bit complicated.
A few closely-related changes:
- Move our definitions of the OpenType spec's "data types" into their
own header file.
- Add definitions for the integer types there too, for completeness.
(Plus Uint16 matches the spec term, and is less verbose than
BigEndian<u16>.)
- Include Traits for the non-BigEndian types so that we can read them
from Streams. (BigEndian<integer-type> already has this.)
- Use the integer types in our struct definitions.
As a bonus, this fixes a bug in Hmtx, which read the left-side bearings
as i16 instead of BigEndian<i16>.
Do more checks at load time, including categorizing the subtables and
producing our own directory of them.
The format for Kern is a little complicated, so use a Stream instead of
manual offsets.
Maxp had the shared fields duplicated, and OS2 embedded each version's
struct in the next. Instead, let's use inheritance to avoid duplicating
shared fields while still allowing them to be directly accessed.
While I'm at it, rename the Maxp and GPOS table structs to just be
VersionX_Y, because they're not ambiguous with anything else.
LibGfx: Rename GPOSHeader to HeaderVersion1_0
Because there's a version 1.1 as well, which we'll eventually want to
support.
